We Built Copilot Cowork Before AI Existed
I've been working on implementing Copilot Cowork since 2011!
The technology we're using with Cowork might be new, but the mindset shift needed to apply its benefits is not. Back in 2011 when I was leading commercial services at a UK university we redesigned our organization with a similar focus to the reasons to consider Cowork today: we consume a lot of specialist effort on tasks that are important but poorly aligned with what those specialists are good at.
The change we designed focused on recognizing that this mundane busywork required specialism in its own right. We established a centralized support team, brought together administrative support resources from multiple departments to give them a team identity, built out digital processes that kept the focus of operational departments on their specialist value, and shifted as much as possible of this organizational procedural activity onto the central team.
We built Cowork. Just with humans.
Today this is exactly the type of shift we are making with AI, but the underlying challenges haven't changed.
1️⃣ Our ability to refocus on where we add the most value is dependent on us acknowledging where we add the least. Many of us feel comfort in busywork that is easy to scratch off our to do lists even while wishing we had more time for the big things. Each AI user today is grappling with what and how to let go of in order to refocus. But this is not a new challenge in the organizational change sphere: creating resource to do work more effectively, and effectively utilizing those new resources, are two distinct steps.
2️⃣ We can only offload to someone (or something) else that which we can describe and model. When it comes to AI advances like Cowork skills, this is still the hurdle we must overcome. For many tasks AI smartness is no longer a barrier, it is solely limited by how well we can describe what we want it to do. This has been exactly the challenge of moving workstreams across organizations for change practitioners for decades.
3️⃣ We need real impact, not performative change. Flashy reorganizations or new technology can paper over cracks in how work is actually being done. We see this today with the rise of tokenmaxxing: give a team a poorly thought out objective and they might just find a way of hitting it without delivering any of the intended benefits. If your goal is to just get everyone using AI without clarity on what they are using it for, then don't be surprised when you end up with malicious compliance instead of real change.
The big difference today with tools like Cowork versus the organizational changes we've seen in the past, is that instead of redesigning our approach to work for large groups, we can now see slightly different redesigns for every individual. This offers new opportunity, but we still need to manage the same risks.
The technology is exciting. But when it comes to the impact on our people, nothing is new here.
🖼️ edited with AI
First posted on Linkedin on 05/29/2026 → View Linkedin Post Here